Documente Academic
Documente Profesional
Documente Cultură
POLONICI MATHEMATICI
94.1 (2008)
The sharp version of a criterion for starlikeness related
to the operator of Alexander
by Rbert Szsz (Corunca)
Abstract. The method of convolution is used to determine a sharp condition for
starlikeness of analytic functions dened in the unit disc U = {z C : |z| < 1} and having
the property f(0) = f
= {f A : Re(zf
(z) + zf
(z)} >
2
6
24
2
= 0.273 . . . , z U,
then f S
(z) + zf
, is
c =
2 ln 2 1
2(1 ln 2)
= 0.629 . . .
In this paper this conjecture will be conrmed.
2. Preliminaries. Let A
0
= {f H(U) : f(0) = 1} and P = {f A
0
:
Re(f(z)) > 0 for all z U}. We will need the following denitions and lem-
mas to prove the main result. Let f(z) =
k=0
a
k
z
k
and g(z) =
k=0
b
k
z
k
be two analytic functions in U. The Hadamard product of f and g is dened
2000 Mathematics Subject Classication: Primary 30C45.
Key words and phrases: operator of Alexander, starlike functions, convolution.
This work was supported by the Research Foundation Sapientia.
[1] c Instytut Matematyczny PAN, 2008
2 R. Szsz
by
(f g)(z) =
k=0
a
k
b
k
z
k
.
For V A
0
, the dual set of V is
V
d
= {g A
0
: (f g)(z) = 0 for all f V and z U}.
Let h
T
A be the function dened by
h
T
(z) =
iT
z
1z
+
z
(1z)
2
1 + iT
, T R.
The following two lemmas are established in [4].
Lemma 1 ([4, p. 94]). The function f A is in the class S
if and only
if
f(z)
z
h
T
(z)
z
= 0 for all z U and T R.
Lemma 2 ([4, p. 23]). The dual set of the class P = {f A
0
: Re(f(z))
> 0, z U} is
P
d
= {f A
0
: Re(f(z)) > 1/2, z U}.
Lemma 3. If the function f : [0, 1] (0, ) is increasing and continu-
ous, then for every a [0, ),
1
0
x
a+1
f(x)
dx
a + 1
a + 2
1
0
x
a
f(x)
dx.
Proof.
1
0
x
a+1
f(x)
dx =
a + 1
a + 2
1
0
x
a
f
_
x
a+1
a+2
_
dx
a + 1
a + 2
1
0
x
a
f(x)
dx.
Lemma 4. If (0, 2) and > 0 then the following identity holds:
(1)
0
x(e
x
+ e
(2)x
)
(
2
+ x
2
)(e
2x
1)
dx + i
0
e
(2)x
e
x
(
2
+ x
2
)(e
2x
1)
dx
=
1
2
+
k=1
e
ik
k +
.
Proof. Consider the function
f(z) =
e
iz
( + z)(e
2iz
1)
, > 0,
where (0, 2) is a xed number. Let N denote the set of natural numbers.
The contour (r, n) is constructed as follows: (r, n) =
1
3
2
4
where
1
(t) = R
n
e
i(t/2)
,
2
(t) = re
i(t+/2)
,
3
(t) = iR
n
+ t(ir iR
n
),
A criterion for starlikeness 3
4
(t) = ir +t(ir iR
n
), t [0, 1], r (0, 1) and R
n
= n+1/2, n N. The
residue theorem implies that
(2)
(r,n)
f(z) dz = 2i
kN, 0<k<n+1/2
Res(f, k).
A computation leads to
(3)
lim
r0
2
f(z) dz = i Res(f, 0),
Res(f, z
k
) = Res(f, k) =
e
ik
2i(k + )
, k N.
We now give a detailed proof for the equality
(4) lim
R
n
1
f(z) dz = 0.
Let (
n
)
n2
be the sequence dened by
n
=
ln n
4
2
(n + 1/2)
, n 2.
We have
(5)
1
f(z) dz
1/2
n
0
f(
1
(t))
1
(t) dt
1/2+
n
1/2
n
f(
1
(t))
1
(t) dt
1/2+
n
f(
1
(t))
1
(t) dt
.
If n > , the inequalities |e
2i
1
(t)
1| e
2(n+1/2) sin(t/2)
1 and
|
1
(t) + | n + 1/2 for t (0, 1/2) imply that
(6)
1/2
n
0
f(
1
(t))
1
(t) dt
1/2
n
0
e
(n+1/2) sin(t/2)
e
2(n+1/2) sin(t/2)
1
n + 1/2
n + 1/2
dt
1
2
1
e
(2)(n+1/2) sin(
n
)
1
n + 1/2
n + 1/2
n
0.
Since lim
t1/2
e
2i
1
(t)
= 1 and lim
n
n
= 0, there is a natural number
n
0
such that |e
2i
1
(t)
1| > 1 for all n > n
0
and all t [1/2
n
, 1/2 +
n
].
4 R. Szsz
Consequently, the second term on the right side of (5) can be estimated by
(7)
1/2+
n
1/2
n
n + 1/2
n + 1/2
e
(n+1/2) sin(t/2)
dt
2
n
n + 1/2
n + 1/2
e
(n+1/2) sin(
n
)
2
n
n + 1/2
n + 1/2
e
2(n+1/2)
n
= 2
n
n + 1/2
n + 1/2
n
n
0.
Finally, the inequalities
|e
i
1
(t)
| e
(n+1/2) sin(t/2)
, |e
2i
1
(t)
1| 1 e
2(n+1/2) sin(t/2)
for t [1/2 +
n
, 1] imply that the third term on the right side on (5) is at
most
(8)
1
1/2+
n
e
(n+1/2) sin(t/2)
1 e
2(n+1/2) sin(t/2)
n + 1/2
n + 1/2
dt
1
2
e
(n+1/2) sin(
n
)
1 e
2(n+1/2) sin(
n
)
n + 1/2
n + 1/2
n
0.
Summarizing, we see that (4) follows from (5)(8). From (2)(4) we obtain
lim
R
n
r0
_
3
f(z) dz +
4
f(z) dz
_
=
1
2
+
k=1
e
ik
k +
,
which is equivalent to
(9)
0
x(e
x
+ e
(2)x
)
(
2
+ x
2
)(e
2x
1)
dx + i
0
e
(2)x
e
x
(
2
+ x
2
)(e
2x
1)
dx
=
1
2
+
k=1
e
ik
k +
.
Lemma 5. Let f : [a, b] [0, ) R be a continuous function and let
g : [0, ) R be an increasing non-constant function. Suppose that:
(i)
f(
1
, x) f(
2
, x)
|
1
2
|q(x) for all
1
,
2
[a, b] and x
[0, ), where q : [0, ) [0, ) is a continuous function and the
improper integral
0
q(x) dg(x) is convergent,
(ii) the function f is dierentiable with respect to the rst variable, and
f/ : (a, b) [0, ) R is continuous,
(iii) the parametric improper integral
0
f
0
f(, x) dg(x),
is dierentiable at every
0
(a, b) and
h
(
0
) =
0
f
(
0
, x) dg(x).
Proof. Let
0
(a, b). There is a > 0 for which I = [
0
,
0
+ ]
(a, b).
According to (i) and (iii) there exists an N > 0 so that g(N) > g(0) and
(10)
M
f
(, x) dg(x)
<
3
and
M
q(x) dg(x)
<
3
for M [N, ) and I.
Condition (i) also implies that
(11)
f(, x) f(
0
, x)
0
dg(x)
N
q(x) dg(x).
The (Lagrange) mean value theorem implies that there is a
(,x)
(0, 1)
such that
f(, x) f(
0
, x)
0
=
f
( +
(,x)
(
0
), x).
Since f/ is uniformly continuous on the compact set I [0, N], there is
a
, then
(, x)
f
(
0
, x)
<
3(g(N) g(0))
for x [0, N].
If |
0
| <
, then | +
(,x)
(
0
)
0
| <
, and so
(12)
f(, x) f(
0
, x)
0
(
0
, x)
<
3(g(N) g(0))
for x [0, N].
Provided that 0 < |
0
| <
h() h(
0
)
0
0
f
(
0
, x)dg(x)
0
_
f(, x) f(
0
, x)
0
(
0
, x)
_
dg(x)
6 R. Szsz
f(, x) f(
0
, x)
0
(
0
, x)
dg(x)
+
f(, x) f(
0
, x)
0
dg(x) +
N
f
(
0
, x) dg(x)
/3 + /3 + /3 = ,
and the proof is complete.
Remark 1. If we put, in Lemma 5, g(x) = [x] where [x] denotes the
integer part of x, then the improper integral
0
f(, x) dg(x) and the se-
ries
n=1
f(, n) are convergent or divergent at the same time. If they are
convergent then
0
f(, x) dg(x) =
n=1
f(, n).
Hence we obtain an analogous lemma for the dierentiability of series.
3. The main result
Theorem 1. If f A and
Re(f
(z) + zf
(z)) >
1 2 ln 2
2(1 ln 2)
for z U,
then f belongs to the class S
(z) + zf
(z) + zf
(z)
1 +
_
> 0, z U,
which can be rewritten as
+ f
+ zf
1 +
P.
According to the representation theorem of Herglotz there is a probability
measure on [0, 2] so that
+ f
(z) + zf
(z)
1 +
=
2
0
1 + e
it
z
1 e
it
z
d(t),
hence, if f(z) = z +
n=1
a
n
z
n
, then
1 +
1
1 +
n=2
n
2
a
n
z
n1
= 1 + 2
n=1
z
n
2
0
e
int
d(t).
A criterion for starlikeness 7
This implies a
n
=
2(1+)
n
2
2
0
e
i(n1)t
d(t) for n N, n 2 and
f(z) = z + 2(1 + )
n=2
z
n
n
2
2
0
e
i(n1)t
d(t).
By Lemma 1, we have to prove that
(13)
f(z)
z
h
T
(z)
z
=
_
1 + 2
n=1
z
n
2
0
e
int
d(t)
_
_
1 + (1 + )
n=1
1 + n + iT
(1 + iT)(n + 1)
2
z
n
_
= 0, z U, T R.
The function dened by the development 1 + 2
n=1
z
n
2
0
e
int
d(t) be-
longs to the class P for every z U and every probability measure . From
Lemma 2 and (13) it follows that f is starlike if and only if the function
g(z) = 1 + (1 + )
n=1
1 + n + iT
(1 + iT)(n + 1)
2
z
n
, z U,
belongs to the dual set P
d
, that is,
Re
_
1 + (1 + )
n=1
1 + n + iT
(1 + iT)(n + 1)
2
z
n
_
>
1
2
, z U, T R.
This inequality is equivalent to
(14) Re
_
1
2(1 + )
+
n=1
1 + n + iT
(1 + iT)(n + 1)
2
z
n
_
> 0, z U, T R.
We will determine
inf
[0,2]
TR
Re
_
n=1
1 + n + iT
(1 + iT)(n + 1)
2
e
in
_
=: m.
If 1/(2(1 + )) = m, then will be the greatest real number for which
(14) is valid. Let us introduce the notation
M(, T) = Re
_
n=1
1 + n + iT
(1 + iT)(n + 1)
2
e
in
_
= Re
_
1 +
1
1 + T
2
n=0
e
in
n + 1
+
T
2
1 + T
2
n=0
e
in
(n + 1)
2
iT
1 + T
2
n=0
e
in
n + 1
+
iT
1 + T
2
n=0
e
in
(n + 1)
2
_
.
8 R. Szsz
The following identities hold for every (0, 2):
n=0
e
in
n + 1
=
1
0
1 t cos
1 + t
2
2t cos
dt + i sin
1
0
t
1 + t
2
2t cos
dt,
n=0
e
in
(n + 1)
2
=
1
0
1
0
1 txcos
1 + t
2
x
2
2txcos
dt dx
+ i sin
1
0
1
0
tx
1 + t
2
x
2
2txcos
dt dx,
and imply
M(, T) = 1 +
1
1 + T
2
1
0
1 t cos
1 + t
2
2t cos
dt
+
T
2
1 + T
2
1
0
1
0
1 txcos
1 + t
2
x
2
2txcos
dt dx
+
T sin
1 + T
2
1
0
t
1 + t
2
2t cos
dt
T sin
1 + T
2
1
0
1
0
tx
1 + t
2
x
2
2txcos
dt dx.
Using the identities
1
0
1 t cos
1 + t
2
2 cos
dt =
1
0
1
1 + t
dt + (1 + cos )
1
0
t(1 t)
(1 + t)(1 + t
2
2t cos )
dt,
1
0
1
0
1 txcos
1 + t
2
x
2
2txcos
dt dx =
1
0
1
0
1
1 + tx
dt dx
+ (1 + cos )
1
0
1
0
tx(1 tx)
(1 + tx)(1 + t
2
x
2
2txcos )
dt dx, (0, 2),
it follows that
M(, T) = 1 +
1
0
1
1 + t
dt +
1
1 + T
2
L
1
(, T),
where
L
1
(, T) := (1 + cos )
1
0
t(1 t)
(1 + t)(1 + t
2
2t cos )
dt
+ T sin
1
0
1
0
t(1 x)(1 t
2
x)
(1 + t
2
2t cos )(1 + t
2
x
2
2txcos )
dt dx
A criterion for starlikeness 9
+ T
2
_
1
0
1
0
t(1 x)
(1 + t)(1 + tx)
dt dx
+ (1 + cos )
1
0
1
0
tx(1 tx)
(1 + tx)(1 + t
2
x
2
2txcos )
dt dx
_
.
If we can prove that
inf
[0,2]
TR
L
1
(, T) = 0,
then we will obtain
inf
[0,2]
TR
M(, T) = 1 + ln 2.
Since L
1
(, T) is a polynomial of the second degree in T with a posi-
tive dominant coecient and L
1
(, 0) = 0, it is sucient to show that the
discriminant of L
1
(, T) satises
L
1
() 0 for all (0, 2). Now,
L
1
() = sin
2
_
1
0
1
0
t(1 x)(1 t
2
x)
(1 + t
2
2t cos )(1 + t
2
x
2
2txcos )
dt dx
_
2
4(1 + cos )
1
0
t(1 t)
(1 + t)(1 + t
2
2t cos )
dt
_
1
0
1
0
t(1 x)
(1 + t)(1 + tx)
dt dx
+ (1 + cos )
1
0
1
0
tx(1 tx)
(1 + tx)(1 + t
2
x
2
2txcos )
dt dx
_
.
Because sup
[0,2]
L
1
() = sup
[0,]
L
1
() we can restrict to [0, ].
Some calculations lead to
L
1
() = 4 cos
2
2
_
sin
2
2
_
1
0
1
0
t(1 x)(1 t
2
x)
(1 + t
2
2t cos )(1 + t
2
x
2
2txcos )
dt dx
_
2
2
1
0
t(1 t)
(1 + t)(1 + t
2
2t cos )
dt
_
1
0
1
0
t(1 x)
(1 + t)(1 + tx)
dt dx
+ (1 + cos )
1
0
1
0
tx(1 tx)
(1 + tx)(1 + t
2
x
2
2txcos )
dt dx
__
.
We will prove the inequality
(15) sin
2
2
_
1
0
1
0
t(1 x)(1 t
2
x)
(1 + t
2
2t cos )(1 + t
2
x
2
2txcos )
dt dx
_
2
2
1
0
t(1 t)
(1 + t)(1 + t
2
2t cos )
dt
1
0
1
0
t(1 x)
(1 + t)(1 + tx)
dt dx
10 R. Szsz
for [/2, ], which implies the claimed condition
L
1
() 0 in the case
[/2, ].
If [/2, ] the function g : [0, 1] R, g(t) = (1 +t)(1 +t
2
2t cos ),
is strictly increasing and so Lemma 3 implies that
2
3
1
0
t
(1 + t)(1 + t
2
2t cos )
dt
1
0
t
2
(1 + t)(1 + t
2
2t cos )
dt,
which is equivalent to
(16)
1
0
t t
2
(1 + t)(1 + t
2
2t cos )
dt
2
3
1
0
1
0
t(1 x)
(1 + t)(1 + t
2
2t cos )
dt dx, [/2, ].
This shows that it is enough to prove
(17) sin
2
2
_
1
0
1
0
t(1 x)(1 t
2
x)
(1 + t
2
2t cos )(1 + t
2
x
2
2txcos )
dt dx
_
2
4
3
1
0
1
0
t(1 x)
(1 + t)(1 + t
2
2t cos )
dt dx
1
0
1
0
t(1 x)
(1 + t)(1 + tx)
dt dx.
This will yield (15) in the case [/2, ].
The inequality of CauchySchwarz implies
(18)
_
1
0
1
0
t(1 x)
(1 + t)
_
(1 + t
2
2t cos )(1 + tx)
dt dx
_
2
0
1
0
t(1 x)
(1 + t)(1 + t
2
2t cos )
dt dx
1
0
1
0
t(1 x)
(1 + t)(1 + tx)
dt dx
for [0, ]. If we prove
(19)
sin
2
0
1
0
t(1 x)(1 t
2
x)
(1 + t
2
2t cos )(1 + t
2
x
2
2txcos )
dt dx
3
1
0
1
0
t(1 x)
(1 + t)
_
(1 + t
2
2t cos )(1 + tx)
dt dx, [/2, ],
then this inequality and (18) imply (17).
A criterion for starlikeness 11
Inequality (19) can be deduced from
t(1 x)(1 t
2
x) sin
2
(1 + t
2
2t cos )(1 + t
2
x
2
2txcos )
3
t(1 x)
(1 + t)
_
(1 + t
2
2t cos )(1 + tx)
for [/2, ] and t, x [0, 1]. The latter inequality is valid because through
direct calculations it can be shown that
sin
2
1 + t
2
2t cos
1
1 + t
and
1 t
2
x
1 + t
2
x
2
2txcos
2
_
3(1 + tx)
,
for [/2, , t [0, 1].
In order to nish the proof, we must also prove that L
1
(, T) 0 for
[0, /2] and T R. To do this, we will apply Lemmas 4 and 5.
We have proved in Lemma 4 the identity
(20)
0
x(e
x
+ e
(2)x
)
(
2
+ x
2
)(e
2x
1)
dx + i
0
e
(2)x
e
x
(
2
+ x
2
)(e
2x
1)
dx
=
1
2
+
k=1
e
ik
k +
.
Let f
1
, f
2
be the functions dened by
f
1
, f
2
: [1/2, 3/2] [0, ) R, f
1
(, x) =
x(e
x
+ e
(2)x
)
(
2
+ x
2
)(e
2x
1)
,
f
2
(, x) =
e
(2)x
e
x
(
2
+ x
2
)(e
2x
1)
.
Since
|f
k
(
1
, x) f
k
(
2
, x)| |
1
2
|
3x(e
2x
+ 1)
((1/2)
2
+ x
2
)
2
(e
2x
1)
=: |
1
2
|q(x),
x (0, ), [1/2, 3/2], k = 1, 2,
and the integral
0
q(x) dx is convergent, the functions f
1
, f
2
satisfy condi-
tion (i) of Lemma 5. Condition (ii) is automatically satised. Since
f
k
(, x)
3x(e
2x
+ 1)
((1/2)
2
+ x
2
)
2
(e
2x
1)
,
x (0, ), [1/2, 3/2], k = 1, 2,
according to a theorem due to Weierstrass the convergence of the integral
0
x(e
2x
+ 1)
((1/2)
2
+ x
2
)
2
(e
2x
1)
dx
12 R. Szsz
implies that the integral
0
f
0
x(e
x
+ e
(2)x
)
(
2
+ x
2
)
2
(e
2x
1)
dx
+ i
0
(
2
x
2
)(e
(2)x
e
x
)
(
2
+ x
2
)
2
(e
2x
1)
dx =
1
2
2
+
k=1
e
ik
(k + )
2
.
In particular, (20) and (21) imply that
0
2xe
x
(1 + x
2
)(e
2x
1)
dx =
1
2
+
n=1
(1)
n
n + 1
=
1
2
+ ln 2,
0
4xe
x
(1 + x
2
)
2
(e
2x
1)
dx =
1
2
+
n=1
(1)
n
(n + 1)
2
=
2
12
1
2
.
Using (20) and (21) for = 1 and the particular cases, we nd that
M(, T) = 1 + ln 2 +
1
1 + T
2
L
2
(, T) +
T
2
1 + T
2
_
2
12
ln 2
_
,
where
(22) L
2
(, T) =
0
x(e
x
+ e
(2)x
2e
x
)
(1 + x
2
)(e
2x
1)
dx
+ T
0
2x
2
(e
(2)x
e
x
)
(1 + x
2
)
2
(e
2x
1)
dx + T
2
0
2x(e
x
+ e
(2)x
2e
x
)
(1 + x
2
)
2
(e
2x
1)
dx.
We have to show that L
2
(, T) > 0 for [0, /2] and T R. L
2
(, T) is a
polynomial of the second degree with respect to T with discriminant
2
() = 4
_
0
x
2
(e
(2)x
e
x
)
(1 + x
2
)
2
(e
2x
1)
dx
_
2
8
0
x(e
x
+ e
(2)x
2e
x
)
(1 + x
2
)
2
(e
2x
1)
dx
0
x(e
x
+ e
(2)x
2e
x
)
(1 + x
2
)(e
2x
1)
dx.
The CauchySchwarz inequality implies
(23)
0
x(e
x
+ e
(2)x
2e
x
)
(1 + x
2
)
2
(e
2x
1)
dx
0
x(e
x
+ e
(2)x
2e
x
)
(1 + x
2
)(e
2x
1)
dx
0
x
1 + x
2
(e
x
+ e
(2)x
2e
x
)
(1 + x
2
)
2
(e
2x
1)
dx
_
2
.
A criterion for starlikeness 13
The inequality
(24)
2
0
x
1 + x
2
(e
x
+ e
(2)x
2e
x
)
(1 + x
2
)
2
(e
2x
1)
dx
0
x
2
(e
(2)x
e
x
)
(1 + x
2
)
2
(e
2x
1)
dx, [0, /2],
and (23) imply that
2
() 0 for [0, /2], which leads to L
2
(, T) 0
for [0, /2] and T R.
On the other hand, the inequality
_
2(1 + x
2
)(e
x
+ e
(2)x
2e
x
) x(e
(2)x
e
x
)
for x (0, ) and [0, /2] implies (24). This inequality is equivalent to
e
()x
_
2(1 + x
2
) + x
_
2(1 + x
2
) x
, x [0, ), [0, /2].
To check this last inequality, it is sucient to consider the case = /2, and
consequently only
e
2
x
_
2(1 + x
2
) + x
_
2(1 + x
2
) x
, x [0, ),
must be proved. This is easily done using the derivative of the function
h : [0, ) R, h(x) = e
2
x
_
2(1 + x
2
) + x
_
2(1 + x
2
) x
.
Theorem 2. The largest value of c for which the condition
(25) f A, Re(f
(z) + zf
(z)) > c, z U,
implies the univalence of the function f, is c =
2 ln 21
2(1ln 2)
.
Proof. According to the proof of Theorem 1, condition (25) implies that
f(z) = z + 2(1 + c)
n=2
z
n
n
2
2
0
e
i(n1)t
d(t)
and
f
(z) = 1 + 2(1 + c)
n=1
z
n
n + 1
2
0
e
int
d(t)
=
_
1 + 2
n=1
z
n
2
0
e
int
d(t)
_
_
1 + (1 + c)
n=1
z
n
n + 1
_
.
14 R. Szsz
Because 1 +2
n=1
z
n
2
0
e
int
d(t) P, Lemma 2 implies that the neces-
sary condition of univalence, f
n=1
z
n
n + 1
_
>
1
2
, z U.
It is simple to prove that condition (26) is equivalent to
1
2(1 + c)
+
n=1
(1)
n
n + 1
0,
which can be rewritten in the form
c
2 ln 2 1
2(1 ln 2)
.
Thus, by Theorem 1 the proof is nished.
4. Integral version of the result. Alexanders operator is dened by
A(f)(z) =
z
0
f(t)
t
dt, z U.
Let R
c
= {f A : Re(f